Target Tracking by Sequential Random Draft Particle Swarm Optimization Algorithm

With the development of economy and the increase of population mobility, the public environment becomes more and more complex. Monitoring system has become an indispensable part of smart city. Target tracking is a key part of monitoring system, tracking is essentially an optimization process, so that, it can be solved by evolutionary algorithms. As evolutionary algorithms with high accuracy and fast convergence, which have attracted increasing attention, particle swarm optimization (PSO) as well as Quantum-behaved Particle Swarm Optimization (QPSO) have been widely used in tracking problem. However, lots of studies have shown that PSO and QPSO all have inherent shortcomings. Falling into local optimum and time consuming make them limited in dealing with tracking applications. For these reason we apply a new random drift particle swarm optimization algorithm (RDPSO) to target tracking. Compared with PSO and QPSO, RDPSO has better global convergence and it is more efficient. Based on traditional PSO-based tracking framework, we propose a sequential RDPSO tracking algorithm. To further improve the performance of the proposed tracking algorithm, we change the particle initialization method, combine the resampling measures in particle filter (PF), and use the Gaussian mixture model to evaluate fitness value. A large number of experimental results show the effectiveness and efficiency of our algorithm, especially for the cases that the background changes greatly, the target is deformed or moves quickly and the camera shakes.